A short introduction to machine learning

Abstract

The field of artificial intelligence aims to develop computer programs that can perform useful tasks like answering questions, recognizing images, and so on. Instead of manually hard-coding all the details of AIs, we specify models with free parameters that are learned automatically from the data they’re given. The dominant paradigm in AI since the 1990s has been machine learning, with deep learning, using neural networks and powerful optimization techniques, becoming the dominant approach in the early 2010s. Machine learning tasks can be supervised, self-supervised, or reinforcement learning, each involving different types of data and objective functions. To apply machine learning techniques to solve real-world tasks, we need to design training setups that are as similar as possible to the real-world task and ensure the AI’s safe behavior by understanding how its skills and motivations will transfer from their training environments to the wider world. – AI-generated abstract.